Deep FusionNet for point cloud semantic segmentation
Many point cloud segmentation methods rely on transferring irregular points into a voxel-based regular representation. Although voxel-based convolutions are useful for feature aggregation, they produce ambiguous or wrong predictions if a voxel contains points from different classes. Other approaches...
| Main Authors: | , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
Springer International Publishing
2020
|
| _version_ | 1826281781186265088 |
|---|---|
| author | Zhang, F Fang, J Wah, B Torr, PHS |
| author_facet | Zhang, F Fang, J Wah, B Torr, PHS |
| author_sort | Zhang, F |
| collection | OXFORD |
| description | Many point cloud segmentation methods rely on transferring irregular points into a voxel-based regular representation. Although voxel-based convolutions are useful for feature aggregation, they produce ambiguous or wrong predictions if a voxel contains points from different classes. Other approaches (such as PointNets and point-wise convolutions) can take irregular points for feature learning. But their high memory and computational costs (such as for neighborhood search and ball-querying) limit their ability and accuracy for large-scale point cloud processing. To address these issues, we propose a deep fusion network architecture (FusionNet) with a unique voxel-based “mini-PointNet” point cloud representation and a new feature aggregation module (fusion module) for large-scale 3D semantic segmentation. Our FusionNet can learn more accurate point-wise predictions when compared to voxel-based convolutional networks. It can realize more effective feature aggregations with lower memory and computational complexity for large-scale point cloud segmentation when compared to the popular point-wise convolutions. Our experimental results show that FusionNet can take more than one million points on one GPU for training to achieve state-of-the-art accuracy on large-scale Semantic KITTI benchmark.The code will be available at https://github.com/feihuzhang/LiDARSeg. |
| first_indexed | 2024-03-07T00:33:58Z |
| format | Conference item |
| id | oxford-uuid:80c17ed9-01ef-486e-bea5-962cc4b56528 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T00:33:58Z |
| publishDate | 2020 |
| publisher | Springer International Publishing |
| record_format | dspace |
| spelling | oxford-uuid:80c17ed9-01ef-486e-bea5-962cc4b565282022-03-26T21:25:37ZDeep FusionNet for point cloud semantic segmentationConference itemhttp://purl.org/coar/resource_type/c_5794uuid:80c17ed9-01ef-486e-bea5-962cc4b56528EnglishSymplectic ElementsSpringer International Publishing2020Zhang, FFang, JWah, BTorr, PHSMany point cloud segmentation methods rely on transferring irregular points into a voxel-based regular representation. Although voxel-based convolutions are useful for feature aggregation, they produce ambiguous or wrong predictions if a voxel contains points from different classes. Other approaches (such as PointNets and point-wise convolutions) can take irregular points for feature learning. But their high memory and computational costs (such as for neighborhood search and ball-querying) limit their ability and accuracy for large-scale point cloud processing. To address these issues, we propose a deep fusion network architecture (FusionNet) with a unique voxel-based “mini-PointNet” point cloud representation and a new feature aggregation module (fusion module) for large-scale 3D semantic segmentation. Our FusionNet can learn more accurate point-wise predictions when compared to voxel-based convolutional networks. It can realize more effective feature aggregations with lower memory and computational complexity for large-scale point cloud segmentation when compared to the popular point-wise convolutions. Our experimental results show that FusionNet can take more than one million points on one GPU for training to achieve state-of-the-art accuracy on large-scale Semantic KITTI benchmark.The code will be available at https://github.com/feihuzhang/LiDARSeg. |
| spellingShingle | Zhang, F Fang, J Wah, B Torr, PHS Deep FusionNet for point cloud semantic segmentation |
| title | Deep FusionNet for point cloud semantic segmentation |
| title_full | Deep FusionNet for point cloud semantic segmentation |
| title_fullStr | Deep FusionNet for point cloud semantic segmentation |
| title_full_unstemmed | Deep FusionNet for point cloud semantic segmentation |
| title_short | Deep FusionNet for point cloud semantic segmentation |
| title_sort | deep fusionnet for point cloud semantic segmentation |
| work_keys_str_mv | AT zhangf deepfusionnetforpointcloudsemanticsegmentation AT fangj deepfusionnetforpointcloudsemanticsegmentation AT wahb deepfusionnetforpointcloudsemanticsegmentation AT torrphs deepfusionnetforpointcloudsemanticsegmentation |